Sign in with
Sign up | Sign in

Graphics Core Next: The Southern Islands Architecture

AMD Radeon HD 7970: Promising Performance, Paper-Launched
By , Igor Wallossek

While the Radeon HD 7970 is the first commercially-available product based on AMD’s Graphics Core Next architecture, the design itself is certainly no secret. In order to give developers some lead time to better exploit its upcoming hardware, Graphics Core Next was exposed in June 2011 at the Fusion Developer Summit. According to Eric Demers, the CTO of AMD’s graphics division, the existing VLIW architecture that was released with Radeon 2000 could still be leveraged for more graphics potential. However, it’s limited in general-purpose computing tasks. Instead of massaging old technology yet again, the company chose to invest in a completely new architecture. 

More compute performance and flexibility are great, but gaming alacrity and visual quality remain the most pertinent responsibilities of high-end desktop graphics hardware. Thus, AMD’s challenge was to create a GPU with a broader focus, simultaneously improving the 3D experience. In order to do that, the company abandoned the Very Long Instruction Word architecture in favor of Graphics Core Next.

The Efficiency Advantage Of Graphics Core Next

AMD’s VLIW architecture is very efficient at handling graphics instructions. Its compiler is optimized for mapping dot product math, which is at the heart of 3D graphics calculations. The design’s weakness is exposed when it has to schedule the scalar instructions seen in more general-purpose applications, though. Sometimes, it turns out that an instruction set, called a wavefront, can’t execute until another wavefront has been resolved. This is called a dependency. The problem is that the compiler can’t change the wavefront queue after it has been scheduled, so precious ALU potential is often wasted as instructions wait until the dependencies are addressed.

Here’s a theoretical example of how a Radeon HD 6970’s VLIW4 SIMD engine and its 16 banks of shader processors (each SP with four ALUs, totaling 64 ALUs per SIMD engine), would handle a wavefront queue that includes dependencies:

As you can see, the VLIW architecture doesn’t handle dependencies ideally. Wavefronts needlessly wait in the queue while free ALUs sit idle.
So, how do you optimize the amount of scalar work you can perform per clock cycle? Enter the Compute Unit, or CU, which replaces the SIMD engines to which we’ve grown accustomed.

Each CU has four Vector Units (VUs), each with 16 ALUs, for a total of 64 ALUs per CU. Thus, the number of ALUs per SIMD/CU is the same. The main difference is that, unlike the shader processors in a SIMD engine, each of the four VUs can be scheduled independently. The CU has its own hardware scheduler that's able to assign wavefronts to available VUs with limited out-of-order capability to avoid dependency bottlenecks.

This is the key to better compute performance because it gives each VU the ability to work on different wavefronts if a dependency exists in the queue:

In our example, the same wavefront queue that took six clock cycles to complete on VLIW4 SIMD engine can be executed in four clock cycles on Graphics Core Next. AMD suggests that Radeon HD 7970 can achieve up to a 7.5x peak theoretical compute performance improvement over the Radeon HD 6970 due to higher utilization. The real-world difference depends on compiler efficiency, and in some compute tasks, the Radeon HD 7970 is barely better than the 6970 on a per-ALU and per-clock basis. We certainly saw lots of variance in our own benchmarks, as you’ll see. But it’s safe to say, based on our synthetic testing, that the Graphics Core Next compute potential exceeds VLIW4.

Deeper Into The Compute Unit

As mentioned previously, the CU replaces the SIMD engines (as named by AMD) we’ve seen since the Radeon 2000 days. We also noted that the CU is composed of four vector units, which, in turn, contain 16 ALUs and a register file each. And we know that the VUs operate independent of each other.

Let’s go into more depth on those VUs, though. Unlike the simplified clock cycle we use in the example above, each VU can process one-quarter of one wavefront per cycle. Equipped with four VUs, each CU can consequently process four wavefronts every four cycles, the equivalent of one wavefront per cycle per CU.

We haven’t yet talked about the CU’s scalar unit, which is primarily responsible for branching code and pointer arithmetic. The vector units could handle those tasks too, but this co-processor’s strength is in offloading scalar work in order to allow the vector units to flex their muscles in more appropriate ways.

Each CU has four texture units tied to a 16 KB read/write L1 cache, which is two times bigger than the VLIW4’s read-only cache. Historically, L1 was only used to read textures. Now, though, they can go back and forth through the same cache.

Ask a Category Expert

Create a new thread in the Reviews comments forum about this subject

Example: Notebook, Android, SSD hard drive

Display all 205 comments.
This thread is closed for comments
Top Comments
  • 42 Hide
    danraies , December 22, 2011 3:49 AM
    cangeliniStart treating your SO super-nice and ask for one for Valentine's Day!


    If I ever find someone that will buy me a $500 graphics card for Valentine's Day I'll be proposing on the spot.
  • 30 Hide
    thepieguy , December 22, 2011 3:32 AM
    If Santa is real, there will be one of these under my Christmas tree in a few more days.
  • 25 Hide
    cangelini , December 22, 2011 4:01 AM
    ZombeeSlayer143WOW!!! I love the conslusion; all of it, which basically is interpretted as "I'm biased towards Nvidia," and trys [desperately] to say don't buy this card! Has the nerve to mention Kepler as an alternative; right, Kepler, as in 1 year away. The GTX580 just got "Radeon-ed" in it's rear. I'm not biased towards either manufacturer, just love to see and give credit to a team of people with passion, vision, and hardwork come together and put their company back on the map, as is shown here today with AMD's launch of the 7970. It's AMD's version of "Tebow Time!!"


    Not at all, actually. It offers great gaming performance, and for everything else...we have to wait, because it's not fully baked yet. They say fools rush in. This thing launched today. We were given less than a week with it. It won't be available to even *buy* for another three weeks. Are you telling me you're ready to crown a paper winner, and you're not biased? Please...
Other Comments
  • 30 Hide
    thepieguy , December 22, 2011 3:32 AM
    If Santa is real, there will be one of these under my Christmas tree in a few more days.
  • 12 Hide
    mi1ez , December 22, 2011 3:40 AM
    Damn, that's a good looking GPU!
  • 11 Hide
    cangelini , December 22, 2011 3:40 AM
    a4mulaFrom a gaming standpoint I fail to see where this card finds a home. For 1920x1080 pretty much any card will work, meanwhile at Eyefinity resolutions it's obvious that a single gpu still isn't viable. Perhaps this will be something that people would consider over 2x 6950, but that isn't exactly an ideal setup either. While much of the article was over my head from a technical standpoint, I hope the 7 series addresses microstuttering in crossfire. If so than perhaps 2x 7950 (Assuming a 449$) becomes a viable alternative to 3x 6950 2GB. I was really hoping we'd see the 7970 in at 449, with the 7950 in at 349. Right now I'm failing to see the value in this card.


    I'll be trolling Newegg for the next couple weeks on the off-chance they pop up before the 9th. A couple in CrossFire could be pretty phenomenal, but it remains to be seen if they maintain the 6900-series scalability.
  • 7 Hide
    cangelini , December 22, 2011 3:41 AM
    thepieguyIf Santa is real, there will be one of these under my Christmas tree in a few more days.


    Hate to break it to you, but there won't be, unless you celebrate Christmas in mid-January.

    Start treating your SO super-nice and ask for one for Valentine's Day!
  • 16 Hide
    Darkerson , December 22, 2011 3:48 AM
    Well I know what I want at tax time :D 
  • 42 Hide
    danraies , December 22, 2011 3:49 AM
    cangeliniStart treating your SO super-nice and ask for one for Valentine's Day!


    If I ever find someone that will buy me a $500 graphics card for Valentine's Day I'll be proposing on the spot.
  • -3 Hide
    Zombeeslayer143 , December 22, 2011 3:56 AM
    No hard feelings to the author...thanks for the review nonethless..
  • 25 Hide
    cangelini , December 22, 2011 4:01 AM
    ZombeeSlayer143WOW!!! I love the conslusion; all of it, which basically is interpretted as "I'm biased towards Nvidia," and trys [desperately] to say don't buy this card! Has the nerve to mention Kepler as an alternative; right, Kepler, as in 1 year away. The GTX580 just got "Radeon-ed" in it's rear. I'm not biased towards either manufacturer, just love to see and give credit to a team of people with passion, vision, and hardwork come together and put their company back on the map, as is shown here today with AMD's launch of the 7970. It's AMD's version of "Tebow Time!!"


    Not at all, actually. It offers great gaming performance, and for everything else...we have to wait, because it's not fully baked yet. They say fools rush in. This thing launched today. We were given less than a week with it. It won't be available to even *buy* for another three weeks. Are you telling me you're ready to crown a paper winner, and you're not biased? Please...
  • 10 Hide
    Zombeeslayer143 , December 22, 2011 4:04 AM
    cangeliniHate to break it to you, but there won't be, unless you celebrate Christmas in mid-January.Start treating your SO super-nice and ask for one for Valentine's Day!


    Dude!!! I'm will be glued to NewEgg via my PDA and PC...I'm sooooo ooohhhh getting two of these!! If they sell for $549, I still think they provide the best value based on current high-end single GPU solutions (e.g., GTX 580, 6970, etc). Nvidia may want to consider a 20%-30% price reduction on the 580's once the 7970's are out.
  • -8 Hide
    Zombeeslayer143 , December 22, 2011 4:15 AM
    ZombeeSlayer143Dude!!! I'm will be glued to NewEgg via my PDA and PC...I'm sooooo ooohhhh getting two of these!! If they sell for $549, I still think they provide the best value based on current high-end single GPU solutions (e.g., GTX 580, 6970, etc). Nvidia may want to consider a 20%-30% price reduction on the 580's once the 7970's are out.


    You have a valid argument there. However, it's not like we need to wait for Windows 8 to see what this GPU is made of, as was the case with Bulldozer. What you see is what we will get give or take 5%?????? Heck, I don't even care for the overclocking potential. If the performance was capped at what is shown here today, I'm certain everyone will be happy. But, of course thats not the case, which brings me back to your point; it's too early, don't get excited, it's just on paper right? LOL...sure buddy...I'm so glad NewEgg was sold out of the GTX580 Classified's the past few weeks; would have regretted that purchase.
  • 6 Hide
    masterofevil22 , December 22, 2011 4:15 AM
    Ahhh....I'll be on the EGG myself and on E of Bay with my 6950 (unlocked) the day this beast this arrives!!!!!!! :D 

    WOW...Go ATI!!!!
  • 15 Hide
    jdwii , December 22, 2011 4:18 AM
    Well Amd you did it, Wow it's been a while since i was able to say that, I'm so impressed with these results everything you guys used to fail on you fixed at your GPU division. I'm not going to lie i did not think Amd would pay it off but they did. All i can say is Nice and well done. You guys beat Nvidia with a 28nnm gpu. I just wonder what's going on at nvidia and why their behind on the 28nm die, And their is rumors that Nvidia wont get their new card out until 2Q of 2012!!


    To bad these cards wont be out this year they would make a lot of money. Oh well parents usually buy them self stuff after christmas i guess.
  • 23 Hide
    aznshinobi , December 22, 2011 4:23 AM
    OMG DAT OVERCLOCKING!
    I can't wait to get the 7850 or 7870 and OC the crap out of it!
  • 10 Hide
    Zero_ , December 22, 2011 4:28 AM
    Overclocking like a baws! It DOES beat a GTX590!
  • 13 Hide
    bavman , December 22, 2011 4:42 AM
    Damn, I didn't see this coming from AMD. This card is a beast...it takes on multi-gpu cards...price is a little high, but compared to the performance it looks like an amazing value.
  • 19 Hide
    hardcore_gamer , December 22, 2011 4:48 AM
    I'm getting one of these for my <30nm build (28 nm GPU and 22nm CPU). I'm callin it the "MOORE'S AR$E"
Display more comments